Human cancers are extremely heterogeneous. Integrative cancer systems biology must develop more comprehensive and incisive methods to interrogate various sources of heterogeneity, and to rationally represent their complex interactions in refined predictive models for individual tumors. One important feature of cancers is the presence of various micro- environmental stresses, such as oxygen depletion, high lactate and extracellular acidosis. These stresses trigger phenotypic changes in cancer cells and directly modulate physiological, metabolic and ultimately clinical phenotypes. Although the importance of these features is well recognized, they are not usually incorporated in clinical decision making because of the significant conceptual and logistic difficulties. In our original R01 grant, we proposed to integrate the influence of lactic acidosis through "genomic gene signatures" based on microarrays. Any one "signature" consists of a set of genes whose expression levels are altered by specific stress in vitro and may therefore be indicative of the micro-environmental effects on the tumor states in vivo. These gene signatures are "common phenotypes" shared by experimental cell cultures and patient tumors to permit the reciprocal flow of information between the contexts: from in vitro mechanistic experimental studies to in vivo molecular and clinical phenotypes of cancer patients, and back. However, in cancer as in other contexts, in vitro cell manipulations cannot fully reflect the complexity and variation seen in human cancers. Therefore, we propose to include additional co-PI, Dr. Lucas to apply sparse statistical factor models and use the expression data of human cancers as "molecular prisms" with which one may dissect and refine the in vitro signatures into in vivo component "factors" or "sub-signatures" which co-vary and interact in complex ways among human cancers. These "sub-signature" factors retain their relationship to the original signature but represent distinct, interacting components of the biological processes represented by the initial signature. Numerical summaries of these dissected in vivo signatures allow us to investigate how these responses are, individually or collectively, linked to oncogenic signaling events in cancers, to DNA copy changes, to other genomic data and to their relationship with clinical outcomes. Variation in the expression levels of many sub-signatures are directly related to variations in the DNA copy number. We will investigate how these changes at the DNA level in cancer cell lines will affect their response to hypoxia and lactic acidosis. The proposed experiments will broaden the scope of our original grant to include comprehensive, quantitative descriptions of the complexity of the interactions of the lactic acidosis and hypoxia responses with genomic and genetic factors, and correlate these interactions with the complex cancer phenotypes that impact patient outcomes. PUBLIC HEALTH RELEVANCE: In this application responsive to the Notice NOT-OD-09-058 titled: NIH Announces the Availability of Recovery Act Funds for Competitive Revision Applications, we propose to apply sparse statistical factor models and use the expression data of human cancers as "molecular prisms" with which one may dissect and refine the in vitro signatures into in vivo component "factors" or "sub-signatures" which co-vary and interact in complex ways among human cancers. These "sub-signature" factors retain their relationship to the original signature but represent distinct, interacting components of the biological processes represented by the initial signature. Numerical summaries of these dissected in vivo signatures allow us to investigate how these responses are, individually or collectively, linked to oncogenic signaling events in cancers, to DNA copy changes, to other genomic data and to their relationship with clinical outcomes.